Imagine knowing how long a battery will last before it even starts degrading. This isn't science fiction; it's the reality researchers are building using the power of artificial intelligence. Lithium metal batteries (LMBs), promising candidates for next-gen energy storage, offer high capacity but suffer from unpredictable lifespans. The long testing times needed to determine a battery's cycle life create a bottleneck in material development. Now, a groundbreaking multi-modal data-driven approach is changing the game. Researchers have developed an 'Automatic Battery data Collector' (ABC) platform that combines a large language model (LLM) and an automatic graph mining tool. This powerful combination extracts vital information about battery materials and performance from scientific papers with remarkable accuracy. The platform analyzes both text and graphs, pulling out details about material composition, operating conditions, and cycle life data. This data fuels machine learning models that accurately predict initial capacity, target cycle capacity, and even the stability of LMBs at different cycle points. These aren't just theoretical predictions; experimental validations confirm the model's real-world applicability. This approach represents a significant step forward, allowing researchers to quickly predict long-term battery performance, potentially accelerating the development of longer-lasting, more efficient LMBs for everything from electric vehicles to portable devices. While the models shine with NCM and LFP cathode batteries, predicting the lifespan of sulfur-based batteries remains a challenge due to the complex chemistry involved. Future research aims to refine the models, incorporating more nuanced data about sulfur-host interactions. The potential for this technology extends beyond batteries. This multi-modal approach could revolutionize data analysis and material discovery in other fields, paving the way for AI-driven breakthroughs in various scientific domains.